Diversify and Conquer: Open-set Disagreement for Robust Semi-supervised Learning with Outliers
This addresses the issue of open-set SSL for machine learning practitioners dealing with real-world data containing unknown classes, though it is incremental as it builds on prior open-set SSL approaches.
The paper tackles the problem of semi-supervised learning (SSL) performance degradation due to outliers in unlabeled data by proposing the Diversify and Conquer (DAC) framework, which uses prediction disagreements among multiple biased models to robustly detect outliers, achieving improved robustness compared to existing methods.
Conventional semi-supervised learning (SSL) ideally assumes that labeled and unlabeled data share an identical class distribution, however in practice, this assumption is easily violated, as unlabeled data often includes unknown class data, i.e., outliers. The outliers are treated as noise, considerably degrading the performance of SSL models. To address this drawback, we propose a novel framework, Diversify and Conquer (DAC), to enhance SSL robustness in the context of open-set semi-supervised learning. In particular, we note that existing open-set SSL methods rely on prediction discrepancies between inliers and outliers from a single model trained on labeled data. This approach can be easily failed when the labeled data is insufficient, leading to performance degradation that is worse than naive SSL that do not account for outliers. In contrast, our approach exploits prediction disagreements among multiple models that are differently biased towards the unlabeled distribution. By leveraging the discrepancies arising from training on unlabeled data, our method enables robust outlier detection even when the labeled data is underspecified. Our key contribution is constructing a collection of differently biased models through a single training process. By encouraging divergent heads to be differently biased towards outliers while making consistent predictions for inliers, we exploit the disagreement among these heads as a measure to identify unknown concepts. Our code is available at https://github.com/heejokong/DivCon.